Why designing effective learning interactions is not easy and how we can do better: Part 1

Speaker
Kenneth R. Koedinger
Professor of Human-Computer Interaction and Psychology, Carnegie Mellon University

When
-

Where
Newell-Simon Hall 1305 (Michael Mauldin Auditorium)

Description

I present an overview of the work of the Pittsburgh Science Learning Center (PSLC), which is midway through its ten years of NSF funding. PSLC’s purpose is to leverage cognitive theory and computational modeling to identify the conditions that cause robust student learning. Helping students learn is not so hard, but finding general methods of instruction that optimize the learning process is a challenge for our time. One aspect of the challenge is the huge space of possible instructional methods. Despite the tendency for educational debates to be framed as though there are two options, like direct instruction vs.constructivism, I will present an analysis that suggests we have trillions of instructional options from which to choose. A second aspect of the challenge starts with the observation that “we don’t know what we know.” Large improvements in instructional effectiveness can be achieved when we use knowledge analysis methods that help us see past our expert blind spots. PSLC researchers are using educational technologies and machine learning methods toward making the current, hard-to-replicate techniques of knowledge analysis (like think aloud) more more reliable and scalable. A third aspect of the challenge, which I will only briefly touch on (this is “Part 1” after all!), is the need to advance our understanding of the vast array of learning mechanisms and strategies that are the causal link between improved instructional methods and robust knowledge acquisition.

I will provide examples from laboratory-quality experiments run, typically with the aid of educational technologies, in the context of math, science, and language learning courses from middle schools to colleges. Results include 1) how a deep knowledge analysis, which suggests algebra learning is like language learning, leads to a transfer result that contradicts the common wisdom that “what you get is what you teach” and 2) how instructional methods that suggest giving students “fewer problems to solve” and “more examples to explain” yield robust learning outcomes in Chemistry, Physics, Algebra, Geometry and, in one study, a reduction in the minority achievement gap.

Speaker's Bio

Kenneth R. Koedinger is a Professor of Human-Computer Interaction and Psychology at Carnegie Mellon University. He has a MS in Computer Science (University of Wisconsin, 1986) and a PhD in Psychology (CMU, 1990). He has authored over 200 papers and has won over 30 major grants. He directs the Pittsburgh Science of Learning Center (see LearnLab.org) and is a co-founder of Carnegie Learning, a company marketing advanced educational technology.